In vitro modeling of tumor spheroid interactions to perfused blood vessels

Tumor angiogenesis, the formation of new blood vessels from existing blood vessels in and around the tumor stroma, is orchestrated by multiple biological factors in the tumor microenvironment, including tumor stromal cells, extracellular matrices, and secreted growth factors. Here, we present the processes to define and optimize the biological conditions for robust interactions between tumor spheroids and engineered blood vessels in a microfluidic organ-on-a-chip device in vitro. Within the device, vascular lumen formation and vessel sprouting in human umbilical vein endothelial cell based engineered blood vessels are observed in a variety of extracellular matrices (collagen, Matrigel, and fibrin), containing metastatic breast tumor cells (MDA-MB-231), and tumor stromal cells (mesenchymal stem cells, and human lung fibroblasts) to show the crosstalk between the tumor spheroids and the perfused blood vasculatures in vitro.


Introduction
Cancer is one of the leading causes of human death in modern society. Tumor angiogenesis, a process of new blood vessel formation from existing blood vessels in and around the tumor stroma has been appreciated as one of the major targets for treating cancer (Folkman, 1971). Tumor blood vessels are morphologically and functionally different from physiological blood vessels, which may affect tumor oxygenation, metabolism, and anti-cancer drug delivery and tumor drug responses (Lee, Pandey, & Popel, 2015). Besides, to better understand and treat tumor metastasis or tumor cell dissemination to distant organs, we need to look into interactions between the tumor mass and nearby blood vessels (Lee et al., 2015).
To model a tumor mass in vitro, tumor spheroids have been used (Fennema, Rivron, Rouwkema, van Blitterswijk, & de Boer, 2013). Tumor spheroids are clusters of tumor cells formed due to the tendency of adherent cells to aggregate when growing in the media suspension (Fennema et al., 2013). Multicellular tumor spheroids have been used to model human solid tumors because of their morphological and biological similarities to in vivo human tumors. Indeed, compared to two-dimensional (2D) tumor cell culture models, three-dimensional (3D) multicellular tumor spheroids could provide more realistic cell-cell, cell-extracellular matrix (ECM) interactions with multiple stromal cells, which are the essential characteristics of the heterogeneous tumor microenvironment (TME) (Chung, Ahn, Son, Kim, & Jeon, 2017;Hirschhaeuser et al., 2010;Ko et al., 2019;Nashimoto et al., 2020). In particular, spheroids models have been useful because they could display morphological and size changes in 3D when the cancer cells' traits are transformed in the ECM under anti-cancer drug treatments (Antoni, Burckel, Josset, & Noel, 2015).
The 3D tumor spheroids models are thus a relevant model for tumor cell growth, migration, and differentiation (Antoni et al., 2015).
In vitro modeling of cancer progression has been emerging (D.H.T. Nguyen et al., 2019;Hassell et al., 2017;Lee, Song, & Chen, 2016). Recently, examining the relationship between blood vessels and tumor spheroids in 3D in vitro devices has attracted attention because the models have offered more realistic environments (e.g., dimensionality, fluid flow (luminal/interstitial), physical deformation, and ECM stiffness) to simulate in vivo human body than conventional cell culture on 2D plastic. For instance, therapeutic efficacies of anti-cancer drugs were assessed in tumor spheroid models under controlled fluid flow in an engineered tumor vascular network (Nashimoto et al., 2020). In this study, the anti-cancer drugs under the flow did not show dose-dependency that has been observed under the static condition. In addition, the proliferation of tumor cells in the spheroids were more accelerated under the flow perfusion (Nashimoto et al., 2020). In another study, differentiation and migration of vascular endothelial cells were observed under pro-angiogenic growth factor treatments, and permeability of the angiogenic sprouts within the collagen matrix was measured (van Duinen et al., 2019). However, it is still challenging to reproduce an appropriate in vitro model system of perfused blood vessels and their crosstalk to solid tumors with various environments (e.g., ECM, stromal cells, growth factors).
In this work, we demonstrate the optimization process to define the biological determinants of successful interactions of tumor spheroids with perfused blood vessels in a microfluidic organon-a-chip device in vitro to mimic vascularized solid tumors in vivo. Within our device, vascular lumen formation and vessel sprouts of the engineered blood vasculatures are observed in a variety of ECM environments containing metastatic tumor cells and supporting stromal cells to show the crosstalk between the tumor spheroids and the blood vasculatures in vitro.

Device fabrication
The Polydimethylsiloxane (PDMS) casting mold was fabricated by following microfabrication method of Polacheck et al (Polacheck, Kutys, Tefft, & Chen, 2019), and the PDMS microfluidic chip device was fabricated using conventional PDMS casting process (Kwak et al., 2016). Briefly, a 10:1 ratio mixture of PDMS polymer and cross-linker curing agent (SYLGARD™ 184 Silicone Elastomer Kit, The Dow Chemical Company, MI, USA) was degassed, cast into the PDMS casting mold, and left in an oven for at least two hours at 80°C. The polymerized PDMS device was punched for inlets and outlets of media reservoir and ECM ports region, and bonded to a glass coverslip with a plasma reacted ion etch (RIE) surface treatment (PE-25 Plasma Cleaner, Plasma Etch Inc., NV, USA). The volume of the ECM cavity in the center of the microfluidic chip device is 20 mm 3 , and the distance between the ECM cavity and the media reservoir is 4 mm. Overall design and dimension of the microfluidic chip device is illustrated in Figure 1.

ECM casting with tumor spheroid
The microfluidic chip device was plasma RIE treated and immersed in 0.1% Poly-L-Lysine PLL (#0413, ScienCell Research Laboratories, CA, USA) for 4 hours and rinsed with DI water three times. Subsequently, 1% glutaraldehyde (G6257, Millipore Sigma, MO, USA) was added to the ECM port and treated for 15 minutes, rinsed with DI water three times, and was soaked overnight in the DI water on an orbital shaker to fully remove any excess glutaraldehyde. The

Vascular cell seeding into devices
Once the ECM materials with tumor spheroids were solidified, the casting needle was removed to leave an empty cylindrical channel completely embedded in the ECM. Afterwards, 1.4 ml of HUVECs were seeded into the channel through the reservoir (cell seeding density at 1 million cells/ml) and allowed to form a monolayer cylindrical vascular channel as shown in the Figure 1b. To apply physiologically pertinent laminar shear stresses (>3 dyne/cm 2 ), the cellseeded microfluidic chip device was placed on a platform rocker with -30° to 30° angle at 3 rpm to generate gravity-driven luminal flow through the vascular channel at 37°C, 5% CO2, 95% humidity. The culture media (EGM-2) in the device was replenished daily.

Immunoblot assays
Proteome Profiler Antibody Array Kits for human angiogenesis factors (R&D systems, MN, USA) were used for reverse western blot. We followed the manufacturer's operation manual to investigate the relative levels of angiogenesis-related proteins in endothelial growth media-2 (EGM-2) and human lung fibroblast conditioned media (HLFCM).

Quantification of spheroid growth pattern and sprouting/invasion length
The normalized area of each growth pattern was analyzed using ImageJ image analysis software (v1.52a, NIH, USA) software with fluorescence color channel splitting, thresholding, and particle analysis module (Kopanska et al., 2015). The average sprouting length was also quantified using ImageJ by manual determination of the furthest distance between the tumor spheroids and the tip of sprouting/invasion.

Statistical analysis
Independent sample populations were compared using unpaired, two-sample t-test with a normal distribution assumption. *P < 0.05 was the threshold for statistical significance. All the error bars depict standard deviation.  I, none of the sprout to the HUVEC channel was found, and the average number of sprouts in the 1: 2 mixture of Matrigel and collagen I was 12.67 ± 6.13 which is slightly more than that in the cases of the 1: 1 and 2: 1 mixture of Matrigel and collagen I which were 12 ± 5.72 and 10.67 ± 4.03 respectively (Table 1 and Figure 3b); However, the average length of sprouts In the case of the 1: 2 mixture of Matrigel and collagen I was 24.75 ± 11.14 µm, which is definitely shorter than that in the cases of the 1: 1 and 2: 1 mixture of Matrigel and collagen I which were 40.48 ± 27.23 µm and 82.81 ± 38.44 µm, respectively (Table 2 and Figure 3c). By contrast, in the Matrigel dominant ECM conditions, the HUVEC vasculatures were all collapsed, although they have relatively more aggressive sprouts, as similar to the results in Figure 2. Consequently, we noticed that none of these ECM conditions tested simultaneously showed both tumor spheroids invasion/sprouts actively and the HUVEC vasculatures are successfully maintained (e.g., lumen formation, media perfusion).

Tumor spheroid-induced angiogenesis
Since we could not simultaneously demonstrate both the tumor spheroids sprout and the perfusable HUVEC vascular lumen in the collagen I and/or Matrigel ECM, as conducted in the above trials, we further explored the fibrin gel embedded ECM microenvironment in the microfluidic chip device. Since it has been reported that integrins in tumor cells and endothelial cells interact with fibrin, facilitating tumor invasion and angiogenesis and synergize proangiogenic growth factors, such as fibroblast growth factor 2 (FGF-2) and vascular endothelial growth factor (VEGF) (Laurens, Koolwijk, & De Maat, 2006;Lee et al., 2015). In addition, we considered additional stromal cell, human lung fibroblasts (HLFs) given the stromal cell support vascular sprouting and tumor invasion (Kim, Chung, & Jeon, 2016;Lee et al., 2015;Lee et al., 2016). We hypothesized that HLFs in fibrin could not only induce more active angiogenesis but also maintain vascular lumens. In this trial, tumor spheroids mixtures of MDA-MB-231 cells (denoted as MDA-231) + HLFs ( Figure 4) were introduced for the interaction between the tumor spheroids and the HUVEC vasculatures. Particularly, for better mimicking of the realistic TME in the human body, we introduced HLFs as a growth factor supplier in the ECM bulk in addition to the tumor spheroids. Interestingly, the mixture of tumor spheroids with MDA-MB-231 cells and HLFs were actively sprouting while the HUVEC vasculature was not collapsed and showed angiogenic sprouting during 6 days of the experiment (Figure 4c-e). The average number and length of the sprouts in this trial were 42 ± 9.42 and 196.08 ± 59.64 µm, respectively. However, although both tumor spheroids and the vasculature were actively sprouted, they had not invaded each other.  Subsequently, we investigated how HLFs induced angiogenesis and maintained vascular lumens. The HLF conditioned media (HLFCM) was prepared as described previously (Lee et al., 2014) to examine the angiogenesis-related protein expression profiles contributed to each experiment. Briefly, when HLFs reached confluence in T175 tissue culture flasks, the normal growth media were replaced with 8 ml serum-free media. After 24 h incubation, the supernatant was centrifuged and filtered through 0.2 mm syringe filters (Corning). The resulting HLFCM were stored in aliquots at -80°C to avoid multiple freeze-thaws. The Figure 6 shows results of the membrane-based antibody array for the determination of the relative levels of angiogenesisrelated proteins in EGM-2 ( Figure 6a) and HLFCM (Figure 6b). Each specimen was introduced to the array membrane and incubated overnight. Then, the relative amounts of 55 cytokines expression levels associated with angiogenesis were detected. Factors in red rectangles represent pro-angiogenic growth factors; and factors in blue rectangles represent anti-angiogenic growth factors ( Figure 6). As the result, we found the pro-angiogenic proteins, Endothelin-1 and VEGF, and the anti-angiogenic proteins, Serpin E1 (PAI-1) were upregulated in both EGM-2 and HLFCM. In addition, the pro-angiogenic proteins, EGF, EG-VEGF, Persephin, bFGF, and the anti-

Discussion
We demonstrated the optimization process of engineered tumor spheroid-induced angiogenesis in a microfluidic organ-on-a-chip device to mimic vascularized breast tumor mass.
We showed that among the ECMs used in the experiments, Matrigel allowed the migration of monoculture of MDA-MB-231 cells or mixed tumor spheroids of MDA-MB-231 cells and MSCs, but significantly disrupted the HUVEC channel by collapsing vascular lumen for unknown reasons ( Figure 2). By contrast, collagen I allowed forming HUVEC lumens, however, the significant migrations of MDA-MB-231 cells in tumor spheroids were not observed ( Figure 3). Interestingly, the fibrin ECM provided the both of forming HUVEC vascular lumens, angiogenesis, tumor invasion more actively than in any case where collagen I or Matrigel was used individually or mixed (Figure 4 and 5). The experimental cases of HLFs embedded in the ECM were showed more active angiogenetic behavior than no cell embedded ECM, and the comparison of the growth media showed that the HLF conditioned media included pro-angiogenic growth factors to encourage angiogenesis under the condition ( Figure 6). These results strongly support the fact that HLFs in ECM have the ability to trigger angiogenesis. Consequently, it was found that a combination of HLFs laden fibrin-ECM, MDA-MB-231 cells and HUVECs mixed tumor spheroids was the most optimal cocktail to induce angiogenesis due to metastatic breast cancer with HUVEC engineered vascular structures among our trials. The reverse western array data revealing secreted factors in HLFs remain further investigation to identify potential new targets to treat tumor angiogenesis.
Feasibility of controlling the experimental complexity is crucial that is required in in vitro models of the complex biomimicking system. The findings of our study represent the potential of the tumor spheroids-vascular crosstalk in the field of cancer research because the 3D tumor spheroids/vascular on-a-chip models provide a more precise in vivo tumor microenvironment (TME) to study tumor growth and development as well as the interactions among tumors, adjacent blood vessels, and extracellular matrices. We believe that the application of these multiple approaches will support obtaining a systematic understanding of developing practical strategies for cancer treatments. Moreover, the use of the system in cancer research will be a potential for the development of more advanced personalized medicine development.
Facility (CNF), a member of the National Nanotechnology Coordinated Infrastructure (NNCI), which is supported by the National Science Foundation (Grant NNCI-1542081).